Support vector machine-based minute-level load curve prediction method

Abstract

The invention discloses a minute-level load curve prediction method based on a support vector machine. The minute-level load curve prediction method comprises the steps of extracting load data of an electrical load and performing data preprocessing; constructing a support vector machine model through the preprocessed data; selecting parameters of the support vector machine model by using a radial basis kernel function; and optimizing parameters of the support vector machine model by using a particle swarm algorithm, inputting the optimized parameters into the support vector machine model, and predicting a load curve based on input and output data of the support vector machine model. The prediction precision of the load curve is improved, and accurate and effective power planning is provided for a power grid.

Description

technical field [0001] The present invention relates to the technical field of load prediction, in particular to a minute-level load curve prediction method based on a support vector machine. Background technique [0002] Load forecasting is the basis of power system planning and design, including forecasting of demand power, maximum load, load rate, load curve, etc. Load curve forecasting plays a very important role in power system engineering. In recent years, load curve forecasting has also become the focus of expert research. So far, the research on load curve forecasting has achieved little results, such as load curve forecasting based on data mining, neural network, non-parametric regression model and other methods. [0003] Nowadays, load curve prediction methods are based on non-parametric models. Although non-parametric regression models are suitable for fitting complex curves, their operations are complex and are not suitable for extended prediction. Moreover, whe...

AI Technical Summary

Problems solved by technology

[0003] Nowadays, load curve forecasting methods are based on non-parametric models. Although non-parametric regression models are suitable for fitting complex curves, they are complex in operation and are not suitable for extension forecasting. Moreover, the "curse of dimensionality" is prone to occur when there are many explanatory variables. There will be phenomena such as a sharp increase in variance, so it is only suitable for low-dimensional spaces

[0004] The medium- and long-term daily load curve prediction problem is transformed into a nonlinear programming problem under linear constraints. It is necessary to predict the daily load characteristics, maximum load and power demand, which is prone to error accumulation and difficult to simulate complex load curves.

[0005] The load curve prediction method based on the holidays of the lunar calendar is only for the daily load curve prediction of the rest days before and after traditional holidays, and is not applicable to normal working days and weekends.

Images